Race-conscious admissions algorithms and the law.
Clinical Snapshot
PICO Framework
| P — Population | Colleges, universities, and educational institutions in the United States subject to the Supreme Court's Students for Fair Admissions (SFFA) ruling |
| I — Intervention | Machine learning (ML)-based admissions systems incorporating various forms of race consciousness (first-order and second-order, at training and predictive phases) |
| C — Comparator | Traditional race-based affirmative action admissions practices as evaluated under the SFFA v. Harvard (2023) Supreme Court decision |
| O — Outcomes | Legal permissibility and normative implications of race-conscious ML admissions algorithms under current US constitutional law; policy calibration possibilities enabled by ML systems |
Bottom Line
This conceptual legal-analytical paper by Chouldechova and Hemel addresses a genuinely important emerging question: how does the US Supreme Court's 2023 SFFA ruling — prohibiting race-based admissions decisions — apply to machine learning systems increasingly used in university admissions? The authors introduce a useful taxonomy distinguishing 'first-order' race consciousness (race as a direct model input) from 'second-order' race consciousness (race used to calibrate or audit model performance), arguing the latter may be legally permissible or even implicitly endorsed by SFFA. The interdisciplinary framing is intellectually rigorous and timely. However, the paper is entirely theoretical: no empirical data on deployed ML admissions systems are presented, no outcomes data on diversity or student success are examined, and no court has yet adjudicated these specific ML questions. The novel taxonomy, while analytically useful, is unvalidated. For Australian clinicians and health system administrators, the paper has no direct relevance — its legal conclusions are US-specific. However, the broader conceptual framework around algorithmic fairness and the distinction between using protected attributes in model training versus prediction may inform thinking about equity in health system resource allocation algorithms. Treat this as a high-quality thought-leadership piece, not as actionable evidence.
Key Findings
P Value: Not applicable — no statistical testing performed
Effect Size: Not applicable — no quantitative effect size reported; this is a legal-normative analytical paper
Primary Outcome: The SFFA v. Harvard (2023) Supreme Court ruling potentially permits — and in some formulations endorses — certain forms of race consciousness in ML-based admissions systems, particularly when race is used at the training phase for calibration rather than as a direct input to individual predictions
Nnt Or Sensitivity: Not applicable — no clinical or diagnostic metrics; the relevant 'measure' is the legal distinction between first-order race consciousness (race as direct predictor input) and second-order race consciousness (race used to calibrate or audit model performance), with the latter argued to be more legally defensible under SFFA
Confidence Interval: Not applicable — no confidence intervals reported
Clinical Application
The conceptual framework is intellectually feasible to apply in institutional policy development, but practical implementation of the proposed ML calibration approaches would require significant technical infrastructure, legal review, and institutional governance capacity. No implementation roadmap or cost analysis is provided. This paper has no direct applicability to Australian healthcare or higher education practice. Australian universities are not subject to the US Equal Protection Clause or SFFA ruling. Relevant Australian frameworks include the Racial Discrimination Act 1975 (Cth), the Higher Education Support Act 2003 (Cth), and TEQSA regulatory requirements. Australian institutions considering ML-assisted admissions would need to consult the Australian Human Rights Commission guidelines on automated decision-making and the Privacy Act 1988 (Cth) regarding sensitive information. The RACGP, PBS, and TGA have no relevance to this paper's subject matter. The conceptual distinction between first-order and second-order race consciousness in ML systems may have some transferable value for Australian policymakers considering algorithmic fairness in admissions or workforce contexts, but legal conclusions must not be transposed. Not directly applicable to clinical populations. Relevant stakeholders include university admissions offices, higher education policymakers, ML system developers, legal counsel for educational institutions, and civil rights advocates in the United States
Abstract
In recent years, colleges and universities have begun to use machine learning (ML) systems to inform admissions decisions. Meanwhile, in the 2023 case Students for Fair Admissions, Inc. v. President and Fellows of Harvard College, the Supreme Court held that colleges and universities may not make admissions decisions "on the basis of race." These parallel developments-the rise of ML in admissions and the fall of race-based affirmative action-will force educational institutions, and ultimately courts, to confront the difficult question of what it means for ML systems to differentiate "on the basis of race." We begin by mapping the Students for Fair Admissions decision onto different uses of race in predictive AI. We distinguish between "first-order" and "second-order" race consciousness at both the training and predictive phases of machine learning, and we argue that each category of race consciousness raises distinct legal and normative issues. We go on to show that the Students for Fair Admissions decision potentially permits-and even endorses-certain forms of race consciousness. Our analysis is grounded in the observation that the process of developing ML-based systems enables policymakers to calibrate decision making algorithms much more precisely and explicitly in response to specific criticisms of race-conscious affirmative action.
References
- 1.Chouldechova, A., & Hemel, D. J. (2026). Race-conscious admissions algorithms and the law. Proceedings of the National Academy of Sciences of the United States of America. https://doi.org/10.1073/pnas.2509764123
This content is for educational purposes for healthcare professionals only and does not constitute clinical advice. Clinical decisions should be based on individual patient assessment, current guidelines, and appropriate specialist consultation. Editorial Standards · Privacy Policy · Terms of Service